Module openai 

OpenAI Chat Completions backend with SSE streaming.

Covers every OpenAI-compatible endpoint: OpenAI itself plus DeepSeek, Groq, Cerebras, xAI, Mistral, Gemini (OpenAI-compat path), Together, Fireworks, OpenRouter, NVIDIA NIM, Moonshot/Kimi, ZAI, MiniMax, Qwen (DashScope compat), HuggingFace router, Vercel AI Gateway, OpenCode Zen, Azure OpenAI, Ollama / LM Studio / vLLM, and any custom INFINITY_OPENAI_BASE_URL proxy.

Structs

Backend

How to reach and authenticate one OpenAI-compatible backend.

Functions

build_body

Build a non-streaming request body (used by tests).

discover_models

Probe an OpenAI-compatible GET {base}/models endpoint.

discovered_ids

Extract model ids from an OpenAI-style GET /models body.

invoke

Stream one completion from an OpenAI-compatible endpoint.